Flight Predictor

During one of my machine learning lecutures a few weeks ago, we were talking about decision trees. My professor was trying to explain the concept of entropy and information gain by illustrating an example of predicting whether flights will be canceled based on certain features (wind, temperature, …etc.).

And that got me thinking. This seems like a great idea! Are there any companies that provide this service of predicting how likely a flight is going to be canceled based on certain factors and past events? Because this seems pretty useful. If you can tell me that there is a 70% chance my flight is going to be canceled based on some circumstances, I’m most likely not going to purchase a ticket for that particular flight anymore.

So I enthusiastically thought about this idea for a couple of hours, but then I suddnely came to a realization that this idea of a flight predictor isn’t as useful or helpful as I thought. Because if you think about how most consumers purchase plane tickets, they usually make a decision weeks or months before the actual flight date. This means that even if I predict a flight has a 90% chance of being canceled, by the time the actual flight date is coming up, circumstances may be drastically different. Just take COVID-19 as an example, the predictor, in February, might say that there’s a mere 5% chance a flight in May is going to be canceled, but we all know that should’ve been more like a 99%. This is an extreme case where circumstances and other determining factors that affects the predictor’s decision change dramatically.

Nevertheless, though this may not be as feasible or useful as I thought, this can still be a pretty cool side project that I can do to practice the materials I learned in class.